QUALITY ANALYST AT XTRACT.IO

Intelligence
needs a
second opinion.

I’m Vignesh. I test the software, workflows and AI decisions that people depend on.

QA automation. Agentic AI 2.0. AiBotPilot.
From the first interaction to the final outcome.

Explore my work Download resume
TAMIL NADU, INDIA Step into my testing lab ↓ OPEN TO QA & AI TESTING ROLES

Functional testing / UI testing / API testing / Automation / Agentic AI validation

A TESTER’S EYE / INTERACTIVE

Looks successful.
Look closer.

A green status is a starting point. Explore a few details I would investigate before calling a feature ready.

AGENT EXECUTION / RESPONSE Completed

Task: extract a company’s revenue from its source report.

01workflow_status"completed"
02stage_status"completed"
03source_report_url""
04revenuenull

Illustrative scenario · no live company data

WHAT I NOTICE

The status and the outcome disagree.

The source report is missing and revenue is null. A completed label does not prove that the extraction succeeded.

WHAT I WOULD VERIFY

  • Required inputs are validated before execution.
  • Missing data produces an accurate failure or review state.
  • The final result satisfies the requested output contract.

01 / SELECTED WORK

Proof is in
the details.

Real areas of responsibility at Xtract.io.
A look at the questions I ask, the flows I test, and the evidence I look for.

01

AGENTIC AI 2.0 / XTRACT.IO

When AI takes action,
quality takes intention.

Validating multi-stage agent workflows for context, execution, approvals and meaningful completion.

Agentic AI validationContext retentionFailure handling
01 Requirement02 Agent execution03 Human approval04 Final output

Each transition deserves a test.

THE QUESTION

Did the task actually succeed?

A workflow can report completion while required inputs are missing or its final output is unusable. I check the relationship between stage status and the actual result.

MY CONTRIBUTION

  • Validate required inputs and failure propagation.
  • Check context after requirement changes.
  • Verify human confirmation and review gates.
  • Compare the final output with the requested schema.

THE QUALITY SIGNAL

Evidence over status labels.

Reproducible test steps, expected versus actual behaviour, and defect evidence that makes the mismatch clear to the team.

02

AIBOTPILOT / XTRACT.IO

From the first prompt
to a publishable bot.

Testing the complete journey through an AI-assisted development workspace: chat, code, compliance, test and onboarding.

Functional & UISession continuityWorkflow gates
Build.Validate.Publish.ONE CONNECTED JOURNEY

THE QUESTION

Does the whole journey hold together?

A successful chat response is one part of the flow. The workspace, editor, generated files and publish actions must remain consistent with it.

MY CONTRIBUTION

  • Verify chat and editor continuity across workspaces.
  • Validate code changes after user confirmation.
  • Check compliance, test and publish gating.
  • Test edit, versioning and onboarding paths.

THE QUALITY SIGNAL

Consistent state, end to end.

Acceptance criteria translated into detailed cases, including negative paths, navigation recovery and changes that span multiple screens.

03

SELENIUM TEST AUTOMATION FRAMEWORK

Repeatable tests.
Readable failures.

A reusable Java automation framework built around page objects, shared utilities and data-driven execution for critical web application flows.

Java + SeleniumTestNG + MavenPOM + Allure
{ } Base test setup[ ] Page objects( ) Reusable actions✓ TestNG → Allure

THE QUESTION

Can another test reuse this reliably?

UI automation needs more than a sequence of clicks. It needs sensible separation, explicit synchronisation and useful failure reporting.

MY CONTRIBUTION

  • Separate page classes and object repositories.
  • Reuse waits, actions and element validations.
  • Handle frames, windows, uploads and alerts.
  • Support data-driven execution and Allure reports.

THE QUALITY SIGNAL

Maintainable test coverage.

Shared behaviours live in reusable utilities, while page objects keep individual flows easier to understand and update.

02 / THE TOOLKIT

Different layers.
One quality mindset.

The tools I use, connected to the work they help me do.

{ } 01

Automation
engineering

Reusable UI checks with clear structure, reliable waits and actionable reports.

  • Core Java
  • Selenium WebDriver
  • TestNG
  • Maven
  • Page Object Model
  • Allure
  • WebDriverWait
↔ 02

API &
data validation

Connect endpoint behaviour to response contracts and the data behind the UI.

  • Postman
  • Swagger
  • REST APIs
  • Oracle SQL
  • JSON validation
  • Data comparison
✳ 03

AI & agent
validation

Evaluate what an AI system says, what it remembers and what it actually does.

  • Agentic AI 2.0
  • AiBotPilot
  • LLM evaluation
  • RAG testing
  • Hallucination detection
  • Source grounding
  • Context retention
⌘ 04

Delivery &
diagnostics

Track work, communicate defects and investigate behaviour across environments.

  • Azure DevOps
  • Git
  • VS Code
  • Eclipse
  • Kibana
  • Grafana
  • Agile / Scrum

HOW WORK MOVES

Azure DevOps is my project and defect management tool—from user stories and acceptance criteria to bug tracking, retesting and closure.

TESTING COVERAGE

Functional · UI · API · Regression · Smoke · Sanity · End-to-end · Agentic AI validation

03 / THE PERSON BEHIND THE TESTS

An engineer’s curiosity.
A tester’s discipline.

VG

ASK. VERIFY. UNDERSTAND.

I’m Vignesh G, a Quality Analyst working at the intersection of software testing, automation and AI.

At Xtract.io, I work on Agentic AI 2.0, AiBotPilot and AI-driven web applications. My testing spans functional behaviour, user interfaces, APIs and the consistency of AI-generated outcomes.

I like getting past “it works” to understand why it works—and where it might fail. That means turning acceptance criteria into meaningful tests, documenting issues clearly, and working with the team to verify the fix.

CURRENT EXPERIENCE

Quality Analyst at Xtract.io

Software Testing · February 2026 – present

ENGINEERING FOUNDATION

B.E. Electronics & Communication

SRM Valliammai Engineering College · 2021 – 2025

English · Tamil · TeluguChennai · Bengaluru · Remote opportunities

BEYOND SOFTWARE TESTING

Ideas I’ve built on.

01 / CAMERA + MACHINE LEARNING

Green Computing-Based
V Ally Shopping Dray

An academic retail automation prototype using intelligent cameras and machine learning for real-time product detection.

Presented at ICCCT 2025 · IEEE publication
02 / COMMUNICATION SYSTEMS

Intelligent Transportation
Using Massive MIMO

An academic traffic-management design exploring Massive MIMO communication to support more reliable transportation systems.

Academic project · April 2024

04 / THE NEXT CHAPTER

Open to opportunities

Let’s make
quality count.

Building a team that cares about reliable software and trustworthy AI?

I’d love to bring my testing mindset to it.

Start a conversation ↗LinkedIn ↗